L. gasseri 1A-TV was grown in de Man Rogosa and Sharpe (MRS) broth and agar plates (Oxoid, Basingstoke, UK) at 37°C for 48 h in anaerobic conditions, using the GasPak™ EZ Anaerobe Pouch System (BD, New Jersey, USA).

In this study, we used seven K. pneumoniae clinical strains from our microbial bank at the MMARLab, previously characterized for their MDR profiles, and selected for their clinical and epidemiological characteristics ( Table 1 ) (Gona et al., 2011, Gona et al., 2019; Mezzatesta et al., 2011). A non-resistant strain was added as control, Kbp7. They comprised the following: (i) Two KPC-producing strains belonging to ST258, isolated from hospitalized patients in two different hospitals in Catania as previously described (Mezzatesta et al., 2011); these strains were isolated from central venous catheter for patient 2 (renamed Kbp1) and from bloodstream for patient 8 (Kbp4 in this study). (ii) Two co-producing NDM-OXA-48 KP belonging to ST101, strain KP730 and KP723 as previously reported (Gona et al., 2019), renamed Kbp2 and Kbp3, respectively; both were isolated from patients of the neonatal intensive care unit of the University Hospital of Catania and belonged to two different epidemiological clusters by the Core Genome MLST (Kbp2 cluster A, Kbp3 cluster C). (iii) Two ESBL-producing strains ST307, isolated from patients 2 and 6 as previously reported (Gona et al., 2011) and named Kbp5 and Kbp6, respectively; Kbp 5 and Kbp6 were responsible for severe urinary tract infections in the renal transplant unit. MLST for K. pneumoniae isolates was performed as previously described (Mezzatesta et al., 2011), and STs were assigned by the MLST Pasteur website (http://www.pasteur.fr/mlst). In addition, minimum inhibitory concentrations (MICs) of amikacin and colistin were performed by the microdilution method interpreted according to EUCAST guidelines v.14.0. (https://www.eucast.org). Fosfomycin MIC was performed by the agar dilution method described by CLSI. All ESBL- and KPC-producing K. pneumoniae were also resistant to colistin. All strains were grown on MacConkey agar (Oxoid, Basingstoke, UK) and aerobically incubated at 37°C overnight.

L. gasseri 1A-TV genomic DNA was extracted using the PureLink™ Genomic DNA Mini Kit (Thermo Fisher Scientific, USA) according to the manufacturer’s instructions, then it was quantified by a Qubit 2.0 fluorometer (dsDNA HS assay, Invitrogen, USA) and quality assessed by a Nanodrop2000 Spectrophotometer (Thermo Fisher Scientific, USA). Genomic libraries were prepared with a Nextera XT DNA Library Prep Kit (Illumina, USA) and sequenced by the Illumina MiSeq platform following the manufacturer’s protocol (Ravi et al., 2018). The raw data were provided as paired-end reads (2 × 250 Read Length) with a 258× coverage and their quality was checked using Trimmomatic (Bolger et al., 2014), while the genome was assembled by SPAdes v3.14.0 through the MEGA annotator pipeline (Bankevich et al., 2012). The genome annotation was performed with Prokka v. 1.13 (Seemann, 2014) ( Supplementary Data Sheet S1 ), whereas rRNA and tRNA gene predictions were performed with RNAmmer v1.2 (Lagesen et al., 2007) and tRNAscan-SE v1.21, respectively (Lowe and Eddy, 1997). Rapid annotation using subsystem technology (RAST) was also applied for microbial genome annotation to identify different subsystems in the genome defined as “a set of functional roles” connected to specific genes of the available annotated genome (Aziz et al., 2008). Furthermore, the database of Cluster of Orthologous Genes (COG) was used for a summary of functional annotation (Galperin et al., 2021).

The L. gasseri 1A-TV genomic reads have been deposited in the National Center for Biotechnology Information (NCBI) SRA database under SUB14624849, with BioProject accession number PRJNA1140194.

Microbial genome mining was obtained using the antiSMASH 7.0 pipeline (antibiotic and Secondary Metabolite Analysis SHell) to predict classic and novel BGCs coding for secondary metabolites (Blin et al., 2023) and BAGEL4 (BActeriocin GEnome minimal tooL) to identify ribosomal synthesized and post-translationally modified peptides (RiPPs) (van Heel et al., 2018). The amino acid sequences of the core peptides identified within the BGCs were further analyzed by the CD search tool.

The inhibitory activity of L. gasseri 1A-TV was firstly tested against seven K. pneumoniae strains shown in Table 1 , using a deferred antagonism test, an assay typically used to investigate the production of bacteriocins and then quantified by the agar spot test (Siroli et al., 2017; Scillato et al., 2021). For the agar spot assay, 5 µL of 1A-TV broth cultures was spotted on the surface of MRS soft agar plates (1.2%) and incubated anaerobically for 48 h at 37°C. An overnight culture of indicator strains (100 μL; approximately 10⁷ CFU/mL) was mixed with Brain Heart Infusion soft agar (0.7%) (Oxoid, Basingstoke, UK) and poured over the plate where lactobacilli were spotted, as previously described (Scillato et al., 2021). Antagonistic activity was assessed by measuring diameters of inhibition zones and expressed as follows: no inhibition (−); diameter between 1 and 3 mm (+); diameter between 3 and 6 mm (++); diameter between 6 and 10 mm (+++); diameter > 10 mm (++++).

Auto-aggregation and co-aggregation activities were determined using our previous protocol with some modifications (Scillato et al., 2021; Vertillo Aluisio et al., 2022). Overnight cultures of L. gasseri 1A-TV and K. pneumoniae were centrifuged at 3,000 g for 15 min at 4°C, and the pellets were washed twice and resuspended in phosphate buffer saline (PBS) (Sigma-Aldrich) to obtain a final concentration of 10⁷ CFU/mL. For auto-aggregation assay, 4 mL of each suspension was incubated at room temperature with a shaker at 200 rpm, whereas for the co-aggregation assay, 2 mL of 1A-TV and 2 mL of each indicator suspension were mixed and incubated in the same conditions. The optical density of bacterial suspensions was measured at 600 nm at T₀ and after 5 h of incubation (T₅). Suspensions were dispensed in a 96-well plate at T₀ and T₅, and OD₆₀₀ readings were attained with a microplate reader (BioTek Synergy™ HTX). For T₅ readings, each sample was centrifuged at 650 g for 2 min to precipitate aggregated bacteria before dispending the 96-well plate.

Auto-aggregation percentage was then calculated as:

Co-aggregation percentage was calculated as:

where AT₀ is the absorbance value at time 0, and AT₅ is the absorbance value after 5 h of incubation.

The co-aggregation percentage of L. gasseri 1A-TV with K. pneumoniae has been considered significant when it is higher than auto-aggregation percentage of each pathogenic strain. Each assay was performed in triplicate.

CFS of L. gasseri 1A-TV was prepared as previously described (Scillato et al., 2021). 1A-TV was inoculated in MRS broth at 37°C for 48 h, then the broth culture was centrifuged at 7,000 g for 30 min at 4°C and the supernatant was filtered with a 0. 22-µm membrane. pH values of CFS were measured by a pH meter (pH 50+DHS Benchtop pH Meter).

¹H-NMR analysis was performed to determine the metabolic profile of L. gasseri 1A-TV CFS. 1A-TV CFS (630 μL; OD₆₀₀ ≈ 1.6) was added to 70 μL of a buffer solution, pH 7.4 [1.5 M K₂HPO₄, 100% (v/v) D₂O, NaN₃ 2 mM, TMSP 5.8 mM]. The final mix (600 μL) was used for the acquisition of the NMR spectra. This analysis was performed using a Bruker IVDr 600 MHz (Bruker BioSpin) operating at the Larmor proton frequency of 600.13 MHz, equipped with a 2H PATXI H/C/decoupling probe 5 mm N and an automatic refrigerated sample changer (SampleJet). The temperature was adjusted to 300 ± 0.1 K with a BTO 2000 thermocouple. The signals were assigned by comparing their chemical shift and multiplicity with the Chenomx software data bank (Chenomx Inc., Canada, ver 8.02). To estimate the reproducibility, 1A-TV CFS NMR analysis was performed three times and MRS free was used as control.

A time-killing assay was performed to evaluate the in vitro inhibitory activity of L. gasseri 1A-TV CFS against K. pneumoniae strains as previously described (Scillato et al., 2021). The kinetic growth of indicator strains in the presence of CFS was tested at different time intervals: 0, 4, 6, 8, and 24 h; viable microbial cell counts (CFU/mL) on Mueller Hinton (MH) agar (Oxoid, Basingstoke, UK) were determined according to the Clinical and Laboratory Standard Institute guidelines (7.17.2 Time-Kill Assay 7.17 tests to assess bactericidal activity, 2024). For the assay, K. pneumoniae indicators were used at 3 × 10⁵ CFU/mL in 10 mL of broth containing 5 mL of cation-adjusted Mueller Hinton broth (BD, New Jersey, USA) and 5 mL of CFS (1:1 ratio). The control growth of indicator strains was assessed by inoculating 10 mL of fresh medium containing MH-MRS broth (1:1). The CFU/mL values of each indicator strain were recorded as log₁₀ reduction of the total count of CFU/mL in the original inoculum and the bactericidal activity was defined as a reduction of at least 99.9% (≥ 3 log₁₀) (approved guideline M26-A, CLSI). This experiment was repeated in triplicate.

The sensitivity of the CFS to enzymatic activity was assayed by trypsin (E. C.3.4.21.4, type II) and proteinase K (E. C. 3.4.21.64) at pH 7.5 (100 mM Tris-HCl buffer). Aliquots of the CFS were incubated (1:1 v/v) with enzyme solutions (1 mg/mL) and their respective controls at 37°C for 2 h under aerobic conditions (Scillato et al., 2021). After proteolytic enzyme treatments, the antibacterial activity of the CFS was performed by antagonism experiments in 96-well plates using the indicator strains at 3 × 10⁵ CFU/mL. After incubation for 24 h at 37°C, the bactericidal effects were estimated by a turbidimetric method with Microplate Reader (BioTek Synergy™ H1) system using OD₆₀₀ for bacterial strains. All experiments were repeated three times.

The crystal violet staining method was used to test biofilm production of K. pneumoniae indicators as previously described with some modifications (Raras et al., 2019; El-Mokhtar et al., 2020). K. pneumoniae ATCC 700603 was used as control for the biofilm production (Ochońska et al., 2021).

Briefly, overnight cultures of K. pneumoniae in LB broth were diluted 1:100 and dispensed in a 96-well plate (200 µL per well). After 24 h of incubation at 37°C, the plate was washed thrice with 200 μL of distilled water and air-dried. Then, 100 μL of 0.2% crystal violet was added to each well, incubated for 20 min, and then discarded. The plate was washed thrice with distilled water, air-dried, then 200 μL of 95% ethanol (Sigma-Aldrich) was added to each well. After 30 min of incubation, the intensity of crystal violet was measured at 570 nm using a microplate reader (BioTek Synergy™ HTX). Based on OD₅₇₀ values, the isolates with OD₅₇₀< 0.120 were considered as weak biofilm producers, those with OD₅₇₀ 0.120–0.240 had intermediate biofilm formation ability, and isolates with OD₅₇₀ > 0.240 were classified as strong biofilm producers, as previously described (Maldonado et al., 2007).

To investigate the effect of L. gasseri 1A-TV CFS on K. pneumoniae biofilm production, only strong biofilm producers were used as indicators. The assay was performed in a 96-well plate as aforementioned with the addition of L. gasseri 1A-TV CFS at scalar concentrations starting from a 1:1 ratio (50% of CFS), then the plate was incubated for 24 h at 37°C. To estimate inhibition of biofilm formation with CFS treatment, the following formula was applied: =

where C denotes the OD₅₇₀ values of wells without CFS treatment, T represents the OD₅₇₀ values of treated wells, and B indicates the OD₅₇₀ values of negative control, as previously described (El-Mokhtar et al., 2020).

To evaluate whether the antibiofilm effect was simultaneously associated with the bactericidal activity of 1A-TV CFS, the same assay was performed in another 96-well plate without crystal violet staining to obtain viable microbial cell counts (CFU/mL). After 24 h of incubation, the cultures were removed from each well, serially diluted and plated on MH agar for determination of viable microbial cell counts. All experiments were performed in at least three independent assays.

Statistical analyses were performed by using GraphPad Prism 8 software (GraphPad Software Inc., San Diego, CA, USA), and results were expressed as mean ± standard deviation (SD) of three independent experiments.

For the co-aggregation assays, ANOVA with Fisher’s least significant difference (LSD) test was applied to determine significant differences. For time-killing assays, the two-way ANOVA test with the Geisser–Greenhouse correction was applied to compare growth curves in treated and untreated conditions. For biofilm inhibition assays, two-way ANOVA with multiple comparisons was used. In all cases, p-value was expressed as follows: * p< 0.05, ** p< 0.01, and *** p< 0.001.

The whole-genome sequencing of L. gasseri 1A-TV showed a single chromosome of 2,018,898 bp in size with a 34.9% GC content, composed of 14 contigs with an N50 value of 352,766. The genome carried 1,951 coding sequences (CDSs), 55 tRNA, and 4 rRNA by RAST annotation. The COG database assigned 1,763 protein-coding genes to 20 different functional categories ( Figure 1 ), such as clustering in transport and metabolism of different macromolecules, energy production and conversion, secondary metabolite biosynthesis, and cell functions. Genome mining using both antiSMASH v.7.0 and BAGEL4 identified two putative bacteriocin biosynthetic gene clusters (BBGCs), namely, BBGC1 (contig_1 RiPP-like) located in contig 1 (from 37,767 to 48,215 nt) that contained two bacteriocin belonging to IIc class, and BBGC2 located in contig 2 (from 10.357 to 31348 nt) carrying the bacteriocin helveticin J of class III. In particular, antiSMASH v.7.0 identified nine putative proteins in the bacteriocin cluster BBGC1, of which six were involved in the bacteriocin biosynthetic process: (i) two class IIc bacteriocins, namely, bacteriocin leader domain-containing protein (ctg1_38; PF10439) and gassericin A-like bacteriocin (ctg1_39, PF10439); (ii) two ABC transporter-related protein (ctg1_35 and ctg1_36; PF00005); (iii) immunity protein (ctg1_40); and (iv) one bacterial extracellular solute-binding protein (ctg_141, PF13416) ( Figures 2A, B ; Supplementary Data Sheet S2 ). Furthermore, the antiSMASH cluster blast showed identical genetic organization of gassericin-S BGC from L. gasseri (BGC0001601) including two bacteriocins (gasA and gasX) and immunity protein (gasI) as shown in Figure 2C (Kasuga et al., 2019). BBGC2 was detected only by BAGEL4 and identified the bacteriocin helveticin J as the core peptide and two bacteriocin ABC transporters (PF00005), which may be involved in the transport of the bacteriocin across the membrane ( Supplementary Data Sheet S3 ).

The effect of antagonistic activity of L. gasseri 1A-TV against seven MDR K. pneumoniae strains evaluated by the deferred antagonism test ( Supplementary Figure 1 ) and quantified by the agar spot test demonstrated a total inhibition of the seven indicator strains with diameters of inhibition zones > 10 mm (++++) for all seven indicator strains ( Figure 3 ).

The auto-aggregation and co-aggregation levels of the indicator strains are shown in Figure 4 . A good ability for auto-aggregation is shown by each strain, with aggregation values ranging between 52% and 66%. Moreover, all K. pneumoniae strains co-aggregated with L. gasseri, but only for K. pneumoniae 1, 2, 3, and 7 was it statistically significant (p< 0.05), with co-aggregation values of 74.6%, 73.6%, 71.9%, and 67.8%, respectively ( Supplementary Table S1 ).

The NMR spectra of CFS L. gasseri 1A-TV identified 25 molecules, mainly metabolites belonging to the families of organic acids, ketones, alcohols, amino acids, monosaccharides, and nucleosides, and concentrations were expressed subtracting the original quantity present in the MRS medium for each component ( Supplementary Table S2 ). Although the 1A-TV metabolic profile included many metabolites involved in the carbohydrate metabolism and amino acid metabolism, organic acids showed the highest concentrations, including ethanol (55.68 mM), lactate (46.62 mM), and acetate (16.61 mM), followed by succinate (7.88 mM) and amino acids like alanine (0.53 mM), valine (0.13 mM), and isoleucine (0.01 mM). Lower levels of certain components compared to the original medium included statistically significant variations for citrate (−2.29 mM), malate (−1.90 mM), and threonine (−0.73 mM) levels.

The results of time-killing assays are summed up in Figure 5 . After 4, 6, and 8 h of incubation, 1A-TV CFS showed a bacteriostatic effect on all ESBL- and KPC-producing strains, with a reduction in bacterial growth of about 1 log₁₀ compared to the initial inoculum. After 24 h of incubation, 1A-TV CFS exerted an extraordinary bactericidal effect against all indicator strains, as the bactericidal activity is defined as a reduction of at least 99.9% (≥ 3 log₁₀) of the total count (CFU/mL) of the original inoculum ( Supplementary Table S3 ). Statistical analysis showed that differences between treated and untreated K. pneumoniae were significant for all strains: p< 0.001 for strains 1, 2, 5, and 7, and p< 0.01 for strains 3, 4, and 6. Furthermore, proteolytic treatment of CFS with proteinase K and trypsin determined a decrease of the antibacterial activity compared to the CFS without proteolytic treatment ( Table 2 ).

To evaluate the ability of 1-ATV CFS to interfere with the K. pneumoniae biofilm formation, the total biomass was evaluated by colorimetric assay using crystal violet. Only K. pneumoniae 1 (ST258, bla _KPC-3, bla _SHV-11, bla _TEM-1, bla _oxa-positive), K. pneumoniae 5 (ST307, bla _{TEM- 1} , bla_{CTX -M-15}-positive), and K. pneumoniae 6 (ST307, bla _{TEM- 1} , bla_{CTX -M-15}-positive) were considered strong biofilm producers with values above the threshold line (OD₅₇₀ > 0.240) and were selected for biofilm inhibition testing ( Figure 6A ).

As shown in Figure 6B , L. gasseri 1A-TV CSF was able to determine a high percentage of biofilm inhibition (from 100% to 95% inhibition) for all K. pneumoniae strains tested at different concentrations, and only at lower concentrations (from 6.25% to 1.56%) was inhibition just slightly reduced with high values from 93% to 86% compared to the free control ( Supplementary Table S4 ). Importantly, no statistical difference was found among the scalar concentrations of CFS. When scalar concentrations of CFS and free control were compared, a statistical difference was found (p< 0.01) for all strains and all concentrations analyzed.

Moreover, viability was assessed after treatment with CFS at scalar concentrations ( Supplementary Figure 2 ): At higher concentrations, CFS inhibited K. pneumoniae growth, despite the higher inoculum compared to time-killing assays, and at concentrations from 12.5% to 1.56% of CFS, all strains had a viable count ranging between 10⁸ and 10¹¹ CFU/mL, but biofilm formation was still inhibited ( Supplementary Table S5 ).